Hyperspectral Remote Sensing-based Study On Estimation Of Moisture Content Of Walnut Leaves

Water is an essential factor in plant growth and development,and leaf water content is an important indicator of plant condition.Continuous water deficit will seriously affect the growth and development,photosynthesis,material transportation,physiological metabolism and yield of walnut.It is of great significance to accurately and effectively monitor the water content of walnut and recover the impact on the tree body by adjusting irrigation practice to ensure the normal production of walnut orchards.In this study,the hyperspectral response characteristics of leaf water changes were analyzed by the naturally occurring water content differences between the leaves of’Xinwen 185’walnut and’Xinxin 2’walnut under conventional management,and the combination of two bands was used to construct The spectral indices were combined to construct a model for leaf moisture content estimation.The spectral indices extracted in 2021 were validated using leaf data from different canopy parts of’Wen185’walnut under irrigation treatment in 2022,and a model for estimating the equivalent water thickness（EWT）of walnut leaves was established.The main results were as follows:（1）Walnut leaf EWT correlated higher with raw spectral reflectance in the short-wave infrared interval overall compared with Relative water content（RWC）and Living fuel moisture content（FLMC）,and the extreme value of EWT correlation was the highest among the three moisture indices.When estimating leaf moisture deficit,it may be more accurate to choose EWT as the moisture metric.（2）The EWT of the vertical profile of walnut leaves showed an increasing trend from the lower to the upper layers,and there were highly significant differences between the EWT of the upper and lower layers of leaves.The reflectance of the leaves from the lower to the upper parts tended to be lower in the shortwave infrared（SWIR）region（1 400-2 500 nm）as a whole.The spectral reflectance of leaves with different water content in SWIR and 400-680nm overall showed that the spectral reflectance decreased with the increase of water content,SWIR reflectance is sensitive to the change of leaf water,is a strong absorption area of water its is a strong absorption valley of water exists near 1 450,1 930 nm.（3）’Xinxin II’walnut EWT had the highest correlation（p=0.893）with the newly constructed spectral index d RSI_{（1 751,2 031）},and the correlation between EWT and first-order spectral indices in the same type of spectral indices were all greater than the original spectral index correlation.The new spectral indices proposed for the three moisture metrics of’Xinwen 185’walnut showed better correlations in the validation of EWT estimation in different canopy parts of walnut,which is complementary to the selection of spectral indices.（4）Compared with the model effect of a single spectral index the model effect constructed using multiple stepwise regression,ridge regression and machine learning methods were significantly improved,and the accuracy and stability of the leaf EWT estimation model constructed by BP neural network algorithm was the best.